`get_turnaround_flags` needs to handle no az motion

[msilvafe]

Currently if the Az is stationary get_turnaround_flags errors out with an error like this:

  File "/so/home/msilvafe/.local/lib/python3.10/site-packages/sotodlib/preprocess/processes.py", line 647, in process
    tod_ops.flags.get_turnaround_flags(aman, **self.process_cfgs)
  File "/so/home/msilvafe/.local/lib/python3.10/site-packages/sotodlib/tod_ops/flags.py", line 199, in get_turnaround_flags
    part_slices_ta_masked = [slice(None, np.where(_ta_flag)[0][0]), slice(np.where(_ta_flag)[0][-1], None)]
IndexError: index 0 is out of bounds for axis 0 with size 0

We need to catch this sort of error and return flags that are all false (or appropriately instantiated). A known satp1 obs that fails in this way is: obs_1706255231_satp1_1111111 and can be used for testing.

[mmccrackan]

For more recent versions of get_turnaround_flags the error output of observations where the Az does not vary significantly is:

"/global/homes/m/mmccrack/so_home/repos/20240819_catchup/sotodlib/sotodlib/tod_ops/flags.py", line 216, in get_turnaround_flags peaks, _ = find_peaks(np.abs(matched), height=peak_threshold, distance=rel_distance_peaks*approx_samps_onescan) "/global/u1/m/mmccrack/so_home/env/soconda_20240903_0.1.4/lib/python3.11/site-packages/scipy/signal/_peak_finding.py", line 941, in find_peaks raise ValueError('distance must be greater or equal to 1')

This occurs because np.ptp is used to find the largest distance between max and min of Az, but this is small for stationary Az observations and may not meet the requirement rel_distance_peaks*approx_samps_onescan be greater than 1. Some might be a few samples bigger though and would not fail.